Salesforce's Agentforce 3 Enhances AI Agent Visibility and Management

Salesforce Agentforce 3 directly addresses a widespread business challenge: gaining clear visibility into what AI agents are actually doing.

Since its initial launch in October 2024, Agentforce has delivered tangible results across multiple industries. Engine managed to reduce customer case handling times by 15 percent, and during the peak of tax season, 1-800Accountant successfully handed off 70 percent of its administrative chat queries to AI.

The significance of this update extends beyond the numbers; it highlights how Salesforce is confronting a major, often unspoken industry problem: businesses are rushing to deploy AI agents without a deep understanding of their operations or how to enhance their performance.

Monitoring your AI agents

The core feature of Agentforce 3 is the Command Center—a centralized control hub for managing your AI workforce. It gives supervisors the tools to analyze agent performance patterns, monitor real-time health indicators like latency and error rates, and pinpoint areas that are effective versus those needing immediate attention.

For organizations that have deployed AI tools but lacked actionable insights, this enhanced oversight could be transformative. The platform logs all agent activity using the OpenTelemetry standard, ensuring seamless integration with existing monitoring tools such as Datadog and Splunk.

The adoption of AI continues to surge. According to upcoming data from the Slack Workflow Index, usage of AI agents increased by 233 percent within just six months. In that same timeframe, around 8,000 organizations signed up to implement Agentforce.

Ryan Teeples, CTO at 1-800Accountant, noted: “During the height of this past tax season, Agentforce independently managed 70% of our administrative chat interactions—providing critical support during one of our busiest periods. But that initial achievement was only the starting point.

“We’ve built a strong deployment framework and are now rolling out new agent-driven workflows and AI automations each week using the latest Agentforce features. With comprehensive observability, we can quickly identify what's effective, make real-time adjustments, and expand our support operations confidently.”

Salesforce Agentforce 3 goes beyond reporting; it actively recommends enhancements. The AI monitors its own interactions, detects conversational trends, and proposes refinements. While slightly self-referential, this capability is invaluable for busy teams that lack the bandwidth to manually review countless automated exchanges.

A solution to the connectivity challenge?

Salesforce is also addressing another common hurdle: connectivity. AI agents can only deliver value if they integrate smoothly with business systems, yet establishing secure and efficient connections has been a persistent challenge.

Agentforce 3 introduces native support for the Model Context Protocol (MCP), which Salesforce aptly compares to “USB-C for AI.” This allows AI agents to connect directly with any MCP-compatible server without requiring custom code, all while adhering to organizational security protocols.

This is where MuleSoft—acquired by Salesforce several years ago—plays a key role, transforming APIs and integrations into agent-friendly resources. Heroku, in turn, manages the deployment and maintenance of custom MCP servers.

Mollie Bodensteiner, SVP of Operations at Engine, added: “Salesforce’s commitment to an open ecosystem, especially through native support for standards like MCP, will be crucial in scaling our AI agent deployments securely.

“We'll be able to link agents directly to essential enterprise systems without custom coding or sacrificing governance. This degree of interoperability gives us the freedom to accelerate adoption while maintaining full oversight of how agents function in our environment.”

Expanding the Salesforce Agentforce ecosystem

Perhaps the most compelling part of this release isn’t what Salesforce developed internally, but the partner ecosystem it's fostering. More than 30 partners—including AWS, Google Cloud, Box, PayPal, and Stripe—have built MCP servers that integrate with Agentforce.

These partnerships deliver more than basic data access. For example, the AWS integration allows agents to process documents, pull information from images, transcribe audio, and even detect key moments in video content. Connections with Google Cloud provide access to Maps, databases, and AI models such as Veo and Imagen.

The healthcare sector stands out as a particularly strong use case.

Tyler Bauer, VP for System Ambulatory Operations at UChicago Medicine, explained: “In healthcare, AI tools must adapt to the intricate and personalized needs of patients and care teams alike.

“Our goal is to automate standard interactions in our patient access center—handling common questions and requests—so our staff can dedicate more time to complex and sensitive patient needs.”

The central question, of course, is whether these advancements will truly help businesses manage their growing fleets of AI agents. Gaining clear insight into AI performance has been a persistent blind spot—many companies know the volume of queries handled by AI but struggle to diagnose specific weaknesses or improvement opportunities.

Adam Evans, EVP & GM of Salesforce AI, stated: “Agentforce 3 will redefine collaboration between humans and AI agents—unlocking new levels of productivity, efficiency, and business transformation.”

While it remains to be seen whether the platform will fully deliver on this ambitious vision, tackling the visibility and control gap is undoubtedly a positive move for companies working to better govern their AI investments.

See also: Huawei HarmonyOS 6 AI agents offer alternative to Android and iOS

Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is co-located with other leading events including Intelligent Automation Conference, BlockX, Digital Transformation Week, and Cyber Security & Cloud Expo.

Explore other upcoming enterprise technology events and webinars powered by TechForge here.